tao-train-deformable-detr
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external datasets in COCO format (JSON annotations and image archives) from local or S3 paths. This data is used as input for training, evaluation, and inference tasks performed by model-specific CLI tools inside Docker containers. While necessary for the skill's primary function, these files represent an untrusted input surface. Ingestion points: Dataset paths defined in SKILL.md and referenced in spec templates. Boundary markers: The skill does not implement explicit boundary markers or instruction-ignoring prompts for the dataset contents. Capability inventory: The skill uses the Bash tool and executes commands within NVIDIA TAO containers. Sanitization: No specific validation or sanitization logic for the content of processed JSON annotations is present in the provided skill files.
Audit Metadata